Chip packaging is normally the final process in the long chain of processes for manufacturing semiconductor integrated circuits. Chip packaging is a multi-disciplinary technology that typically involves many steps. The technology is critically important because it has a direct impact on chip performance and reliability as well as the performance and reliability of electronic devices in which the chips are incorporated. “Packaging” as used herein encompasses any of various conventional techniques of preparing a chip (also called a “die”) for actual use in an electronic device. In many instances, packaging involves, basically, encapsulating the die in a manner that seals and protects the die from the external environment and provides the required external electrical connections (called “I/O” connections) from the die to other circuitry. Packaging also can facilitate the conduction of heat away from the die during use. Other types of packaging may simply involve mounting the die on a substrate or carrier, with which the die makes the required I/O connections, without forming a discrete capsule around the individual die beforehand. Since die encapsulation consumes space, this latter packaging method is typically used in applications in which size is critical, such as electronic watches, hearing aids and other medical devices, cellular phones and other personal communication equipment, and high-speed microprocessors.
The current disclosure is directed in general to the making of electrical connections of I/O bond-pads on a die to corresponding I/O bond-pads on a substrate, carrier, lead-frame, or the like. These electrical connections provide the required I/O connections of the die to the world outside the die.
Wire bonding has, for many years, been a “workhorse” technology for making electrical connections between the I/O bond-pads on the die and the I/O bond-pads on the package or other die-mounting substrate. However, wire bonding has several disadvantages. First, it typically is performed serially, pad-by-pad, which is inherently slow and thus decreases throughput. Second, as the number of I/O connections to an integrated circuit (e.g., memory or microprocessor chip) has increased, increasingly larger numbers of I/O bond-pads on the die are required. Providing a larger number of such pads without excessively increasing the size of the chip usually requires a corresponding decrease in the pitch (i.e., a finer pitch) of I/O bond-pads on the die, wherein “pitch” is the center-to-center distance between adjacent bond-pads, and a “finer” pitch is a decreased center-to-center distance. These factors, as well as other factors, have increased the difficulty and decreased the reliability of using wire bonds, which has led to much interest in alternative methods of making I/O connections to individual dies.
Key alternative methods are derived from the so-called “flip-chip” technology. Flip-chip involves the assembly of a die to a substrate or carrier in a face-down manner by using electrically conductive bumps formed on the I/O bond-pads of the die. (“Face-down” means that the die surface on which the circuit layers are formed actually faces the substrate to which the die is attached. Wire-bonding, in contrast, is performed on face-up dies). Flip-chip methods made their debut in the mid-1960s but did not achieve widespread utilization for many years largely because wire-bonding was the norm. With the advent of extremely complex integrated circuits requiring large numbers of I/O connections, flip-chip methods became attractive because, for example: (a) they can be performed more reliably with a finer bond-pad pitch than wire-bonding; (b) they can be performed at lower cost than wire-bonding; (c) all the I/O connections to a die can be formed simultaneously rather than serially as in wire bonding; and (d) the reliability of devices formed using flip-chip methods has been proven. Currently, flip-chip components are predominantly semiconductor devices such as integrated circuits, memories, and microprocessors; however, flip-chip methods are also being used increasingly with other types of devices as well, such as passive filters, detector arrays, and MEMs devices. Flip-chip is also termed “direct chip attach” (abbreviated DCA), which is perhaps a more descriptive term because the die is attached directly to the substrate, carrier, or the like by the conductive bumps. DCA has allowed, in certain instances, elimination of a conventional “package” entirely.
Among the various conventional flip-chip methods, the most common technique is the “solder-bump” technique that forms a small, individual solder “bump” (typically a roughly spherical bump) on each of the I/O bond-pads of the die. Certain aspects of this structure are shown in FIGS. 4(A)-4(B), depicting a die 10, a substrate 12, I/O bond-pads 14 on the substrate, and I/O bond-pads 16 on the die. The solder-bump technique usually commences, before the wafer (containing multiple dies) is diced, with formation of “under-bump metallization” (“UBM”) 18 on the I/O bond-pads 16 of the dies 10. Forming the UBM 18 follows removal of an insulating aluminum oxide layer on the I/O bond-pads 16 to expose elemental aluminum, and includes formation of a series of metal layers on defined regions of the exposed aluminum. Thus, the UBM 18 defines and limits the respective regions on the bond-pads to be wetted by solder in a subsequent step. Solder is deposited on the UBM regions 18 by evaporation, electroplating, screen-printing of solder paste, or needle-deposition, for example. After deposition of the solder, the wafer is heated to re-flow the solder into individual solder “balls” 20 having roughly spherical shape (FIG. 4(A)). The wafer is then diced into “bumped dies” 22. Individual bumped dies 22 are placed accurately on respective substrates 12 or carriers (generally termed “substrates”), as shown in FIG. 4(A). Each such assembly is heated to form the solder connections 24 between the die 10 and the I/O bond-pads 14 on the substrate, as shown in FIG. 4(B). After the die 10 is soldered, “underfill” (not shown, but usually an epoxy adhesive) typically is added between the die 10 and the substrate 12.
Despite certain advantages of the UBM-solder bump technology summarized above, it has certain limitations. One limitation is posed by the minimum practical size of the individual solder balls (exemplary size of each ball being 70-100 μm high and 100-125 μm in diameter), which inherently imposes limits on the minimum size of the I/O bond-pads and thus on the achievable fineness of bump pitch. Another limitation is the need to remove the wafers from the wafer-fabrication operation in order to form the solder balls, which imposes a throughput-lowering and cost-increasing disruption in the overall chip-manufacturing process. Yet another limitation is posed by the fact that the solder used for forming the balls contains lead, which is a potential source of contamination that, if not rigorously contained, can lead to device failure.